1 /* ----------------------------------------------------------------------
2 * Project: CMSIS DSP Library
3 * Title: arm_power_f16.c
4 * Description: Sum of the squares of the elements of a floating-point vector
5 *
6 * $Date: 23 April 2021
7 * $Revision: V1.9.0
8 *
9 * Target Processor: Cortex-M and Cortex-A cores
10 * -------------------------------------------------------------------- */
11 /*
12 * Copyright (C) 2010-2021 ARM Limited or its affiliates. All rights reserved.
13 *
14 * SPDX-License-Identifier: Apache-2.0
15 *
16 * Licensed under the Apache License, Version 2.0 (the License); you may
17 * not use this file except in compliance with the License.
18 * You may obtain a copy of the License at
19 *
20 * www.apache.org/licenses/LICENSE-2.0
21 *
22 * Unless required by applicable law or agreed to in writing, software
23 * distributed under the License is distributed on an AS IS BASIS, WITHOUT
24 * WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
25 * See the License for the specific language governing permissions and
26 * limitations under the License.
27 */
28
29 #include "dsp/statistics_functions_f16.h"
30
31 #if defined(ARM_FLOAT16_SUPPORTED)
32
33
34 /**
35 @ingroup groupStats
36 */
37
38
39
40 /**
41 @addtogroup power
42 @{
43 */
44
45 /**
46 @brief Sum of the squares of the elements of a floating-point vector.
47 @param[in] pSrc points to the input vector
48 @param[in] blockSize number of samples in input vector
49 @param[out] pResult sum of the squares value returned here
50 */
51 #if defined(ARM_MATH_MVE_FLOAT16) && !defined(ARM_MATH_AUTOVECTORIZE)
52
53 #include "arm_helium_utils.h"
54
arm_power_f16(const float16_t * pSrc,uint32_t blockSize,float16_t * pResult)55 void arm_power_f16(
56 const float16_t * pSrc,
57 uint32_t blockSize,
58 float16_t * pResult)
59 {
60 int32_t blkCnt; /* loop counters */
61 f16x8_t vecSrc;
62 f16x8_t sumVec = vdupq_n_f16(0.0f);
63
64
65 blkCnt = blockSize;
66 do {
67 mve_pred16_t p = vctp16q(blkCnt);
68
69 vecSrc = vldrhq_z_f16((float16_t const *) pSrc, p);
70 /*
71 * sum lanes
72 */
73 sumVec = vfmaq_m(sumVec, vecSrc, vecSrc, p);
74
75 blkCnt -= 8;
76 pSrc += 8;
77 }
78 while (blkCnt > 0);
79
80 *pResult = vecAddAcrossF16Mve(sumVec);
81 }
82 #else
83
arm_power_f16(const float16_t * pSrc,uint32_t blockSize,float16_t * pResult)84 void arm_power_f16(
85 const float16_t * pSrc,
86 uint32_t blockSize,
87 float16_t * pResult)
88 {
89 uint32_t blkCnt; /* Loop counter */
90 _Float16 sum = 0.0f16; /* Temporary result storage */
91 _Float16 in; /* Temporary variable to store input value */
92
93 #if defined (ARM_MATH_LOOPUNROLL) && !defined(ARM_MATH_AUTOVECTORIZE)
94
95 /* Loop unrolling: Compute 4 outputs at a time */
96 blkCnt = blockSize >> 2U;
97
98 while (blkCnt > 0U)
99 {
100 /* C = A[0] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1] */
101
102 /* Compute Power and store result in a temporary variable, sum. */
103 in = *pSrc++;
104 sum += in * in;
105
106 in = *pSrc++;
107 sum += in * in;
108
109 in = *pSrc++;
110 sum += in * in;
111
112 in = *pSrc++;
113 sum += in * in;
114
115 /* Decrement loop counter */
116 blkCnt--;
117 }
118
119 /* Loop unrolling: Compute remaining outputs */
120 blkCnt = blockSize % 0x4U;
121
122 #else
123
124 /* Initialize blkCnt with number of samples */
125 blkCnt = blockSize;
126
127 #endif /* #if defined (ARM_MATH_LOOPUNROLL) */
128
129 while (blkCnt > 0U)
130 {
131 /* C = A[0] * A[0] + A[1] * A[1] + ... + A[blockSize-1] * A[blockSize-1] */
132
133 /* Compute Power and store result in a temporary variable, sum. */
134 in = *pSrc++;
135 sum += in * in;
136
137 /* Decrement loop counter */
138 blkCnt--;
139 }
140
141 /* Store result to destination */
142 *pResult = sum;
143 }
144 #endif /* defined(ARM_MATH_MVEF) && !defined(ARM_MATH_AUTOVECTORIZE) */
145
146 /**
147 @} end of power group
148 */
149
150 #endif /* #if defined(ARM_FLOAT16_SUPPORTED) */
151
152
